# Seedream 5 Lite - Core.Today AI API

> Seedream 5.0 lite: image generation with built-in reasoning, example-based editing, and deep domain knowledge. Supports multi-reference generation with up to 14 images and sequential batch generation.

- **Provider**: ByteDance
- **Model ID**: bytedance/seedream-5-lite
- **Category**: Image Generation
- **Credits**: 82 per image (2K/3K same price; sequential auto bills 81 × max_images)
- **Speed**: Fast
- **Quality**: High

## Features
- Example-based editing — show a before/after pair and the model applies the same transformation
- Logical reasoning for spatial relationships, physics, and processes
- Deep domain knowledge across architecture, science, health, and design
- Multi-image blending with up to 14 reference images
- Sequential batch generation for storyboards, brand packages, and character sheets
- Accurate multi-language text rendering — wrap text in double quotes

## Use Cases
- Scientific diagrams and labeled cross-section illustrations
- Typographic posters and marketing materials with accurate text
- Photorealistic renderings from floor plan sketches
- Story scenes and character variations in one batch request
- Complex edits demonstrated by example instead of described in words

## API Endpoint

Base URL: https://api.core.today/v1
Create Prediction: POST /predictions
Get Status: GET /predictions/{job_id}

## Authentication
Header: X-API-Key: YOUR_API_KEY

## Input Parameters

### Required
- **prompt**: string - Text prompt (max 4000 chars). Keep under 600 English words for best results. Wrap desired in-image text in double quotes

### Optional
- **image_input**: array (default: []) - 1-14 input images for single or multi-reference generation
- **size**: string (default: 2K) - Image resolution: 2K (2048px) or 3K (3072px) — same price
  Options: 2K, 3K
- **aspect_ratio**: string (default: match_input_image) - Aspect ratio. 'match_input_image' follows the input image's ratio
  Options: match_input_image, 1:1, 4:3, 3:4, 16:9, 9:16, 3:2, 2:3, 21:9
- **sequential_image_generation**: string (default: disabled) - 'disabled' generates one image; 'auto' lets the model emit multiple related images (billed per image)
  Options: disabled, auto
- **max_images**: integer (default: 1) - Max images when sequential mode is 'auto' (1-15). Input + generated cannot exceed 15
- **output_format**: string (default: png) - Output image format
  Options: png, jpeg

## Examples

### Reasoned Scene Generation
Complex contemporary scene rendered in a classical painting style
```json
{
  "model": "bytedance/seedream-5-lite",
  "input": {
    "prompt": "A Renaissance-style oil painting of a modern-day farmer's market. A woman in a hoodie examines heirloom tomatoes under a striped awning, lit like a Vermeer interior. Warm golden Dutch Golden Age light, every detail unmistakably contemporary",
    "size": "2K",
    "aspect_ratio": "16:9"
  }
}
```

### Typographic Poster
Multi-line text rendering with precise typography
```json
{
  "model": "bytedance/seedream-5-lite",
  "input": {
    "prompt": "A large-format typographic poster for a jazz festival. At the top in bold condensed sans-serif: \"BLUE NOTE SESSIONS\" in deep navy. Below in elegant script: \"Summer 2026 — Central Park, New York.\" A golden saxophone silhouette runs along the right edge",
    "size": "2K",
    "aspect_ratio": "2:3"
  }
}
```

## Response Format
```json
{
  "job_id": "abc123",
  "status": "pending | processing | completed | failed",
  "result": "URL or data (when completed)"
}
```

## Usage Flow
1. POST /predictions with model and input -> receive job_id
2. GET /predictions/{job_id} -> poll until status is completed or failed
3. Result contains output URL(s)

## Tips
- Use natural language sentences, not keyword lists — the model reasons through full descriptions
- Wrap any text you want rendered in the image in double quotes ("...")
- For hard-to-describe edits, provide a before/after example pair plus your new image as inputs
- When editing, state what must stay unchanged: 'Replace the hat with a crown, keeping pose and expression'
- State your use case ('Design a logo for a gaming company') for better-targeted results
- 3K costs the same as 2K — use 3K when you need larger output

## Documentation
https://replicate.com/bytedance/seedream-5-lite